Enhancing Search in E-Commerce Platforms

This thesis presents the design and implementation of an advanced product search engine for PrestaShop based e-commerce platforms, addressing the critical challenge of helping customers efficiently discover products in online retail environments. The system integrates Elasticsearch with modern semantic search capabilities to overcome the limitations of traditional keyword-based search systems. The research develops a modular, microservices based architecture combining lexical and semantic retrieval methods. At its core, the system employs Elasticsearch for robust indexing and BM25 ranking, augmented with Sentence Transformer embeddings to enable semantic understanding of user queries beyond literal keyword matching. A FastAPI backend serves as the bridge between user interactions and search functionality, while custom JavaScript solutions seamlessly integrate advanced features into PrestaShop's existing interface. Comprehensive evaluation demonstrates significant improvements over baseline PrestaShop search functionality. BM25 implementation achieves perfect recall (1.0) and substantially improved precision (0.7081) compared to the baseline (0.5244), with near-optimal ranking quality (nDCG@10: 0.9982). Semantic search further enhances precision to 0.7243, successfully capturing user intent and related concepts even when queries use different terminology than product descriptions. The system enables effective cross-linguistic matching. For example, retrieving "calice" or "coppa da vino" when users search for "bicchiere vino." The implementation includes practical enhancements such as AJAX-based autocomplete suggestions, comprehensive faceted filtering, and user interaction analytics for continuous improvement. The resulting scalable, flexible, and maintainable architecture provides e-commerce platforms with enterprise-grade search capabilities, directly improving product discovery, reducing search abandonment, and enhancing conversion rates.